[NTLUG:Discuss] Samba Server comments = performance hit

[Richard Geoffrion]

Kenneth Loafman wrote:
> Chris Cox wrote:
>   
>> <SNIP SNIP SNIP>
>> Acording to John Terpstra... it's a huge hit.  Just fyi.  I think
>> he knows what he's talking about.
>>
>> There's a reason why some distributions moved away from a highly
>> commented smb.conf file.
>>     
>
> If that's the case then they need to rethink their config strategy.  If 
> they are doing more than checking for a changed timestamp, then they are 
> doing serious overkill.  Even cron, which checks multiple files and dirs 
> once a minute does not have a high overhead.
>   

From: http://lists.samba.org/archive/samba/2002-November/056815.html
==================
Someone named John T. wrote:

/My first exposure to real problems with excessively sized smb.conf files
//was in a site that had 800 MS Windows clients. They had a full smb.conf
//config history in the file. It was 156Kb in size. At that time, samba read
//smb.conf (as it does now) every 20 seconds, except now we only read the
//actual file if the last change is more recent than when smbd started.
/
/Anyway, at this site the system load with all clients idle was 30! Work it
//out:
/
/800 x 156Kb x 3/60 = 6.2MB/sec of file system I/O without doing a thing.
//By optimizing the smb.conf file this dropped the load to nothing.  The
//resulting file was just under 10Kb in size.
/
/So you see, this radically taught me that config files are no place for
//documentation. That should be done in a separate file. SWAT optimizes the
//smb.conf file for minimum size, only writing out parameters that are not
//at default./

============


Based on this information and information from time spent in #Samba and 
#Samba-technical, typical comments in an smb.conf file will pose no 
great performance risk.  

I would, however, discourage one from storing one's autobiography in a a 
production smb.conf file.

-- 
Richard